The DelPhi Multiprocessor Inference Machine
In 1986 I proposed a new principle of executing Horn Clause programs with low overhead on multiple processors. This method does not involve the sharing of memory or the copying of computation state between processors. Under certain conditions, the method has the remarkable property of executing a given program in the minimum time theoretically required. Such optimal operation is not always possible, but performance of the implemented system is such as to render it of practical use. This paper describes the experience of implementing the method on a network of processors and of executing programs according to the method. This paper is more a retrospective than a tutorial, and so readers are referred to previous papers for introductory material and technical details.
KeywordsMultiple Processor Performance Graph Deterministic Problem Proof Tree Variable Binding
Unable to display preview. Download preview PDF.
- 1.H. Alshawi and D.B. Moran. The Delphi model and some preliminary experiments. Proc 5th Conf Symp Log Prog (ed Kowalski and Bowen), MIT Press, 1578–1589.Google Scholar
- 2.W.F. Clocksin, 1987. Principles of the DelPhi parallel inference machine. Computer Journal 30(5), 386–391.Google Scholar
- 4.W.F. Clocksin and H. Alshawi, 1986. A method for efficiently executing Horn Clause programs using multiple processors. Technical Report CCSC-3, SRI International (Cambridge Computer Science Centre).Google Scholar
- 6.C.S. Klein, 1991. Exploiting or-parallelism in Prolog using multiple sequential machines. PhD dissertation. Reprinted as Technical Report 216, Computer Laboratory, University of Cambridge.Google Scholar
- 7.D.H.D. Warren, 1987. OR-parallel execution models of Prolog. Technical Report, Department of Computer Science, University of Manchester.Google Scholar
- 8.K.L. Wrench, 1990. A distributed and-or parallel prolog network. PhD dissertation. Available in summary form as Technical Report 212, Computer Laboratory, University of Cambridge.Google Scholar